Surface plasmon resonance

Transcription

1 Surface plasmon resonance P. Anton van der Merwe 1. INTRODUCTION 3 2. PRINCIPLES AND APPLICATIONS OF SURFACE PLASMON RESONANCE Principles Applications What SPR is good for What SPR is not good for GENERAL PRINCIPLES OF BIACORE EXPERIMENTS A typical experiment Preparation of materials and buffers Monitoring the Dips 9 4. LIGAND Direct versus indirect immobilisation Covalent immobilisation A general approach Choice of chemistry Prepare the protein Pre-concentration Amine coupling Regeneration Adjusting the immobilisation conditions Non-covalent immobilisation (ligand capture) Using an existing strategy Developing a new strategy Activity of immobilised ligand Control surfaces Re-using sensor chips ANALYTE Purity, activity and concentration

2 5.2. Valency Refractive index effect and control analytes Low molecular weight analytes QUALITITATIVE ANALYSIS; DO THEY INTERACT? Positive and negative controls Qualitative comparisons using a multivalent analyte QUANTITATIVE MEASUREMENTS Affinity Concepts Experimental design Data analysis Controls Non-linear Scatchard Plots Kinetics Concepts Experimental design Data analysis Controls Stoichiometry Thermodynamics Activation energy CONCLUSION APPENDIX Physical basis of SPR Samples and buffers Guidelines for preparing samples and buffers for use on the BIAcore Standard buffers Additional information REFERENCES 41!!" #$%&'()*+,*-*&./

3 12. GLOSSARY FIGURE 1. ANALYSIS OF DIPS. 44!0".#1)*/ Table 1. BIAcore instruments currently available (January 1999) Table 2. Sensor chips available for use on the BIAcore Table 3. Covalent coupling chemistry Table 4. Techniques for ligand capture Table 5. Trivial causes of complex binding kinetics Table 6. Distinguishing some non-trivial causes of complex kinetics Introduction Since the development almost a decade ago (1,2) of the first biosensor based on surface plasmon resonance (SPR), the use of this technique has increased steadily. Although there are several SPRbased systems (3-5), by far the most widely used one is the BIAcore (1,2), produced by BIAcore AB, which has developed into a range of instruments (Table 1). By December 1998 over 1200 publications had reported results obtained using the BIAcore. It is likely that it would be even more widely used were it not for its high cost and the pitfalls associated with obtaining accurate quantitative data (5-8). The latter has discouraged many investigators and led to the perception that the technique may be flawed. This is unjustified because the pitfalls are common to many binding techniques and, once understood, they are easily avoided (4,8,9). Furthermore, the BIAcore offers particular advantages for analysing weak macromolecular interactions, allowing measurements that are not possible using any other technique (4,10). This Chapter aims to provide guidance to users of SPR, with an emphasis on avoiding pitfalls. No attempt is made to describe the routine operation and maintenance of the BIAcore, as this is comprehensively described in the BIAcore instrument manual - 3 -

4 (see Section 9.3). Although written for BIAcore users, the general principles will be applicable to experiments on any SPR instrument. 2. Principles and applications of Surface Plasmon Resonance Principles The underlying physical principles of SPR are complex (see Section 9.1). Fortunately, an adequate working knowledge of the technique does not require a detailed theoretical understanding. It suffices to know that SPR-based instruments use an optical method to measure the refractive index near (within ~300 nm) a sensor surface. In the BIAcore this surface forms the floor of a small flow cell, nl in volume (Table 1), through which an aqueous solution (henceforth called the running buffer) passes under continuous flow (1-100!L.min -1 ). In order to detect an interaction one molecule (the ligand) is immobilised onto the sensor surface. Its binding partner (the analyte) is injected in aqueous solution (sample buffer) through the flow cell, also under continuous flow. As the analyte binds to the ligand the accumulation of protein on the surface results in an increase in the refractive index. This change in refractive index is measured in real time, and the result plotted as response or resonance units (RUs) versus time (a sensorgram). Importantly, a response (background response) will also be generated if there is a difference in the refractive indices of the running and sample buffers. This background response must be subtracted from the sensorgram to obtain the actual binding response. The background response is recorded by injecting the analyte through a control or reference flow cell, which has no ligand or an irrelevant ligand immobilized to the sensor surface. One RU represents the binding of approximately 1 pg protein/mm 2. In practise >50 pg/mm 2 of analyte binding is needed. Because is it very difficult to immobilise a sufficiently high density of ligand onto a surface to achieve this level of analyte binding, BIAcore have developed sensor - 4 -

5 surfaces with a nm thick carboxymethylated dextran matrix attached. By effectively adding a third dimension to the surface, much higher levels of ligand immobilisation are possible. However, having very high levels of ligand has two important drawbacks. Firstly, with such a high ligand density the rate at which the surface binds the analyte may exceed the rate at which the analyte can be delivered to the surface (the latter is referred to as mass transport). In this situation, mass transport becomes the rate-limiting step. Consequently, the measured association rate constant (k on ) is slower than the true k on (see Section 7.2). A second, related problem is that, following dissociation of the analyte, it can rebind to the unoccupied ligand before diffusing out of the matrix and being washed from the flow cell. Consequently, the measured dissociation rate constant (apparent k off ) is slower than the true k off (see Section 7.2). Although the dextran matrix may exaggerate these kinetic artefacts (mass transport limitations and re-binding) they can affect all surface-binding techniques Applications This section outlines the applications for which SPR is particularly well suited. Also described are some applications for which it is probably not the technique of choice. Of course, future technical improvements are likely to extend the range of applications for which the SPR is useful What SPR is good for Evaluation of macromolecules Most laboratories studying biological problems at the molecular or cellular level need to produce recombinant proteins. It is important to be able to show that the recombinant protein has the same structure as its native counterpart. With the possible exception of enzymes, this is most easily done by confirming that the protein binds its natural ligands. Because such interactions involve multiple residues, which are usually far apart in the primary amino acid sequence, they require a correctly folded protein. In the absence of natural ligands monoclonal antibodies (mabs) that are known to - 5 -

6 bind to the native protein are an excellent means of assessing the structural integrity of the recombinant protein. The BIAcore is particularly well suited to evaluating the binding of recombinant proteins to natural ligands and mabs. Setting up an assay for any particular protein is very fast, and the data provided are highly informative Equilibrium measurements (affinity and enthalpy). Equilibrium analysis requires multiple sequential injections of analyte at different concentrations (and at different temperatures). Because this is very time-consuming it is only practical to perform equilibrium analysis on interactions that attain equilibrium within about 30 min. The time it takes to reach equilibrium is determined primarily by the dissociation rate constant or k off ; a useful rule of thumb is that an interaction should reach 99% of the equilibrium level within 4.6/k off seconds. High affinity interactions (K D < 10 nm) usually have very slow k off values and are therefore unsuitable for equilibrium analysis. Conversely, very weak interactions (K D >100!M) are easily studied. The small sample volumes required for BIAcore injections (<20!L) make it feasible to inject the very high concentrations (>500!M) of protein required to saturate low affinity interactions (11). Equilibrium affinity measurements on the BIAcore are highly reproducible. This feature and the very precise temperature control makes it possible to estimate binding enthalpy by van t Hoff analysis (12). This involves measuring the (often small) change in affinity with temperature (Section 7.4). Although not as rigorous as calorimetry, much less protein is required Kinetic measurements The fact that the BIAcore generates real-time binding data makes it well suited to the analysis of binding kinetics. There are, however, important limitations to kinetic analysis. Largely because of mass transport limitations it is difficult to measure accurately k on values faster than about 10 6 M -1 s -1. This upper limit is dependent on the size of the analyte. Faster k on values can be measured with - 6 -

7 analytes with a greater molecular mass. This is because the larger signal produced by a large analyte allows the experiment to be performed at lower ligand densities, and lower ligand densities require lower rates of mass transport. For different reasons measuring k off values slower than 10-5 s -1 or faster than ~1 s -1 is difficult. Because the BIAcore is easy to use and the analysis software is user-friendly, it is deceptively easy to generate kinetic data. However obtaining accurate kinetic data is a very demanding and time-consuming task, and requires a thorough understanding of binding kinetics and the potential sources of artefact (Section 7.2) Analysis of mutant proteins It is possible using BIAcore to visualise the capture of proteins from crude mixtures onto the sensor surface. This is very convenient for analysing mutants generated by site-directed mutagenesis (13,14). Mutants can be expressed as tagged proteins by transient transfection and then captured from crude tissue culture supernatant using an antibody to the tag, thus effectively purifying the mutant protein on the sensor surface. It is then simple to evaluate the effect of the mutation on the binding properties (affinity, kinetics, and even thermodynamics) of the immobilised protein. This provides the only practical way of quantifying the effect of mutations on the thermodynamics and kinetics of weak protein/ligand interactions (15) What SPR is not good for High throughput assays. The fact the BIAcore can only sample can be analysed at a time, with each analysis taking 5-15 min, means that it is neither practical nor efficient for high throughput assays. Automation does not solve this problem because the sensor surface deteriorates over time and with re-use. Blockages or air bubbles in the microfluidic system are also common in long experiments, especially when many samples are injected

8 Concentration assays. The BIAcore is also unsuitable for concentration measurements, because these require the analysis of many samples in parallel, including the standard curve. A second problem is that, for optimal sensitivity, concentration assays require long equilibration periods Studying small analytes. Because the SPR measures the mass of material binding to the sensor surface, very small analytes (M r <1000) give very small responses. The recent improvements in signal to noise ratio have made it possible to measure binding of such small analytes. However a very high surface concentration of active immobilised ligand (~1 mm) is needed, and this is difficult to achieve. Furthermore, at such high ligand densities accurate kinetic analysis is not possible because of mass-transport limitations and re-binding (Section 7.2). Thus only equilibrium analysis is possible with very small analytes, and then only under optimal conditions. This assessment may need to be revised as and when future improvements are made in the signal to noise ratio. 3. General principles of BIAcore experiments 3.1. A typical experiment A typical SPR experiment involves several discrete tasks. "# Prepare ligand and analyte. "# Select and insert a suitable sensor chip. "# Immobilise the ligand and a control ligand to sensor surfaces. "# Inject analyte and a control analyte over sensor surfaces and record response. "# Regenerate surfaces if necessary. "# Analyse data

9 While the ligand and analyte could be almost any type of molecule, they are usually both proteins. This chapter focuses on the analysis of protein-protein interactions Preparation of materials and buffers BIAcore experiments are frequently disrupted by small air bubbles or other particles passing through the flow system. Usually these can be flushed out and the experiment repeated, wasting only time and reagents. Occasionally, however, the damage is irreversible necessitating the replacement of an expensive sensor chip or integrated fluidic cartridge. This can be minimised by following a few simple rules (Section 9.2.1) Monitoring the Dips The output from the photo-detector array that is used to determine the surface plasmon resonance angle ($ spr, Section 9.1) can be viewed directly as 'dips'. The current BIAcore documentation does not describe how to view and interpret dips, and so this information is supplied here (Protocol 1). Protocol 1. Normal and abnormal 'dips' Viewing the dips 1. Enter the service mode on the BIAcore control software by simultaneously pressing the control, alt, and s keys. 2. When the dialog box appears requesting a password ignore this and press the enter or return key. An additional Service menu will appear on the menu bar. 3. Select View dips from this menu. 4. A graph appears similar to the one in Figure 1 showing the amplitude of light reflected off the sensor surface (Reflectance) measured over a small range of angles. The angle of the minimum reflectance ( $ spr ) is calculated by fitting a curve to all this data, thereby enabling $ spr to be measured at a far greater resolution than the resolution at which the data are actually collected

10 Normal dip The important feature of a normal dip is its depth (Figure 1, Dip1). Generally it bottoms out at a Reflectance of ~ When the refractive index above the sensor surface increases (e.g. because of protein binding), the dip shifts to the right while maintaining its shape and depth and (Dip 2). Abnormal dip There are two main types of abnormal dip. 1. The shallow dip (Dip 3). The $ spr is measured along a section of the sensor surface rather than at a single point. Refractive index heterogeneity on the surface gives heterogenous $ spr values. When averaged the result is a shallow dip that does not reach below RUs. Heterogeneity can be the result of differences in the amount of material immobilised along the surface, in which case the dip is usually slightly shallow, or the result of small air bubbles or particles in the flowcell, in which case the dip is very shallow. 2. No dip (Dip 4). A large change in the refractive index beyond the instrument dynamic range (Table 1) will shift the $ spr so much that no dip is evident. Usually this is the result of air in the flow cell. While slight shallowness of the dip is acceptable, and is common after coupling large amounts of protein, more severe abnormalities should not be ignored. Attempts should be made to return dips to normal by flushing the flow cells with buffer and/or regenerating the sensor surfaces. If this is ineffective a new sensor surface should be used. 4. Ligand 4.1. Direct versus indirect immobilisation The most challenging step when setting up SPR experiments is immobilising of one of the proteins (the ligand) to the sensor surface without disrupting its activity. Immobilisation can either be direct, by covalent coupling, or indirect, through capture by a covalently coupled molecule. The major advantage of direct covalent immobilisation is that it can be used for any protein provided that it is reasonably pure (>50%) and has a pi > 3.5. However, it has three important drawbacks

11 Firstly, because they usually have multiple copies of the functional group that mediates immobilization, proteins are coupled heterogeneously and sometimes at multiple sites. Secondly, direct coupling often decreases or completely abrogates binding to analyte. And thirdly, directlycoupled proteins are difficult to regenerate (see below). Indirect immobilisation has the disadvantage that it can only be used for proteins that have a suitable binding site or tag for the covalently coupled molecule. However, it has four important advantages, which make it the method of choice in most cases. Firstly, proteins are seldom inactivated by indirect coupling. Secondly, the protein need not be pure. It can be captured from a 'crude' sample. Thirdly, all the molecules are immobilised in a known and consistent orientation on the surface. Finally, using appropriate buffers it is often possible to dissociate selectively the non-covalent ligand/analyte bond, thereby enabling the ligand surface to be re-used a process termed regeneration Covalent immobilisation A general approach Instead of replicating the detailed protocols in the BIAcore literature (see Section 9.3), I will suggest a general approach (Protocol 2) to covalent coupling, discussing several aspects in detail. Protocol 2. An approach to covalent coupling of a protein 1. Select the coupling chemistry (Section 4.2.2). 2. Prepare the protein (Section 4.2.3). 3. Optimize the pre-concentration step (Section 4.2.4). 4. Couple the protein (Section 4.2.5). 5. Evaluate the activity of the immobilised protein (Section 4.5). 6. Establish conditions for regeneration (Section 4.2.6). 7. Adjust the immobilization conditions (Section 4.2.7)

12 Choice of chemistry There are three main types of coupling chemistry, which utilize, respectively, amine (e.g. lysine), thiol (cysteine) or aldehyde (carbohydrate) functional groups on glycoproteins (Table 3). All covalent coupling methods utilize free carboxymethyl groups on the sensor chip surface. They can therefore be used for any of the sensor chips that have such carboxymethyl groups (See Table 2). If the protein to be immobilised has a surface-exposed disulphide or a free cysteine, ligand-thiol coupling is probably the method of choice. Failing this, amine coupling should be tried in the first instance. If amine coupling inactivates the protein (as assessed by ligand and/or mab binding), aldehyde coupling can be attempted, provided that the protein is glycosylated. Detailed protocols are available from BIAcore for all coupling techiques (see Section 9.3). Only amine coupling is described here in some detail Prepare the protein Only a modest amount (5-10!g) of protein is needed. The major requirement is that the protein is pure and has a high level of activity. Because direct coupling is relatively indiscriminate, all protein in the preparation will be coupled. Thus if the preparation is contaminated by other proteins or is partially active, the level of binding observed will be proportionally decreased. Because the protein needs to be diluted into the pre-concentration buffer (Section 4.2.4) to a final concentration of 20-50!g/ml, the stock should be fairly concentrated (> 0.5 mg/ml). If the protein is in solutions that are strongly buffered or contain high salt concentrations, primary amine groups or sodium azide, then these must either be dialysed out or much larger dilutions (1:100) made. In the latter case the protein will need to be at higher concentrations (>2 mg/ml) Pre-concentration

13 The purpose of pre-concentration is to concentrate the protein to very high levels (>100 mg/ml) within the dextran matrix, thereby driving the coupling reaction. Without pre-concentration far higher concentrations of protein would need to be injected to get equivalent levels of coupling. Preconcentration is driven by an electrostatic interaction between the negatively charged carboxylated dextran matrix and positively charged protein. To this end the protein is diluted into a buffer with a low ionic strength (to minimize charge screening) with a ph below its isoelectric point or pi (to give the protein net positive charge). However amine coupling is most efficient at a high ph because activated carboxyl groups react better with uncharged amino groups. Thus the highest ph compatible with pre-concentration is determined empirically (Protocol 3). Electrostatically-bound protein should dissociate rapidly and completely when injection of running buffer resumes, both because the proteins net positive charge will decrease and because electrostatic interactions will be screened by the high ionic strength of the running buffer. Incomplete dissociation suggests that the interaction was not purely electrostatic, perhaps because of the binding, at low ph, of a denatured form of the protein. Protocol 3. Determining the optimum pre-concentration conditions for a protein 1. Dilute the protein to a final concentration of 20-50!g/mL (final volume 100!L) into ph 6.0, 5.5, 5.0, 4.5 and 4.0 pre-concentration buffers (see Section 9.2.2). 2. Start a manual BIAcore run using a single flow cell (flow-rate 10!L/min) 3. Inject 30!L of each sample, beginning at ph 6.0 and working down. 4. If no electrostatic interaction is observed continue with protein samples diluted in ph 3.5 and 3.0 pre-concentration buffers. 5. Use the highest ph at which >10000 RU of protein binds electrostatically during the injections. 6. Check that all the bound protein dissociates after the injection. If not it suggests that not all the protein was bound electrostatically and that the protein denatures irreversibly at that ph

14 Amine coupling There is a standard protocol for amine coupling (Protocol 4). The first step is to activate the carboxymethyl groups with N-hydroxysuccinimide (NHS), thus creating a highly reactive succinimide ester which reacts with amine and other nucleophilic groups on proteins. The second (coupling) step is to inject the protein in pre-concentration buffer, thereby achieved high protein concentrations and driving the coupling reaction. The third (blocking) step, blocks the remaining activated carboxymethyl groups by injecting very high concentrations of ethanolamine. The high concentration of ethanolamine also helps to elute any non-covalent bound material. The final regeneration step is optional. When a protein is coupled for the first time it is advisable not to include any regeneration step. If it is included and the protein has poor activity it will not be clear whether covalent coupling or regeneration was responsible for disrupting the protein. The structural integrity of the protein should be evaluated (Section 4.2.6) before regeneration is attempted. Once regeneration conditions have been established these can be added on to the coupling protocol. Protocol 4. Amine coupling Reagents 120!L of protein at 20-50!g/ml in suitable pre-concentration buffer (Protocol 3) 120!L of EDC (0.4 M) mixed 1:1 with NHS (0.1 M). Make up just before coupling 120!L of ethanolamine/hcl 1 M, ph !L of regeneration solution (if regeneration conditions known) Procedure 1. Establish pre-concentration conditions (Protocol 3) 2. Set flow-rate to 10!L/min

15 3. Activation: inject 70!L of EDC/NHS 4. Coupling: inject 70!L of protein 5. Blocking: inject 70!L of ethanolamine 6. Regeneration: inject 30!L of regeneration solution (if regeneration conditions known, see Protocol 5). 7. View the 'dips' (Protocol 1) to confirm that the immobilisation is homogeneous Regeneration Once a covalently immobilised protein has been shown to be active with respect to binding its natural ligand or a monoclonal antibody (Section 4.4), regeneration can be attempted. A general approach to establishing regeneration conditions can be used (Protocol 5). The goal here is to elute any noncovalently bound analyte without disrupting the activity of the ligand. Regeneration allows surfaces to be re-used many times, saving both time and money. However establishing ideal regeneration conditions can be a very time-consuming, and in many cases impossible, task. Thus it may be more cost-effective to opt for imperfect or no regeneration, using new sensor surfaces instead. Protocol 5. Establishing regeneration conditions Regeneration solutions "# These fall into four main classes: divalent cation chelator, high ionic strength, low ph, high ph (Section 9.2.2). "# If the interaction is likely to be dependent on divalent cations try buffers with EDTA. "# If the monomeric interaction is weak, try high ionic strength buffers. "# Otherwise start with a low ph buffer. "# If this is without effect, try high ph buffer. Approach 1. Covalently couple ligand to the surface. 2. Make up enough analyte for several injections

16 3. Inject analyte over the immobilised ligand and measure the amount of binding. 4. Inject selected regeneration buffer for 3 min, and measure the amount of analyte that remains bound after this. 5. If the decrease is <30% switch to a different class of regeneration buffer and return to step If the decrease is >30% but <90%, try repeated injections. If this fails to elute >90% select a stronger regeneration buffer of the same type and return to step If this fails, try a different class of buffer and return to step When >90% of bound analyte is eluted, return to step 3, using identical analyte and injection conditions. If binding post-regeneration remains >90% of binding before regeneration, the regeneration conditions may be adequate. 9. If residual activity is <90%, select a different class of regeneration buffer. If residual binding is very low it may be necessary to return to step 1, starting with a new surface. 10. If two different types of buffer give partial elution, try using both sequentially Adjusting the immobilisation conditions. Since the level of ligand coupling achieved is unpredictable it is usually necessary to modify the initial protocol in order to achieve the desired immobilization level. It is also often necessary to create several surfaces on the same chip with different levels of coupling. The best way to achieve different levels of coupling is to change the duration of the activation step, by varying the volume of NHS/EDC injected. The level of coupled ligand varies in proportion with the duration of the activation step. Thus, a two-fold reduction in activation period will usually lead to a two-fold reduction in coupling. It is possible using an option in the inject command to couple ligand simultaneously in multiple flow cells, varying only the length of activation step Non-covalent immobilisation (ligand capture) There are two requirements for non-covalent or indirect immobilization of a ligand (henceforth called 'ligand capture'). First, it must be possible to obtain or create a sensor surface that can capture a ligand. Ideally, it should be possible to regenerate this surface so that repeated capture is possible

17 Second, the ligand needs to have a suitable binding site or modification that allows it to be captured Using an existing strategy A number of established techniques are available for ligand capture (Table 4). When expressing a recombinant protein to be used in SPR studies it is advisable to consider adding a suitable domain or peptide motif so that one of these techniques can be used. The precise choice of tag will depend on whether the tag is needed for purification and what other uses are envisaged for the ligand. "# If it is envisaged that the ligand is to be used as an analyte in SPR studies it should be monovalent or, if multivalent (Fc- and GST-chimeras), the tag should be readily removable. "# If the ligand is to be used for structural studies, the tag should also be removable. "# If the ligand is to be used to probe for binding partners on cells or tissues it should be multivalent, or it should be possible to make it multivalent. "# If the ligand is to be captured from crude mixtures (i.e. after expression without purification) the capturing agent needs to have a high affinity and to be highly specific. For example, CD4 (14) and anti-human IgG 1 (13) mabs have been shown to be suitable for this purpose Developing a new strategy The widespread availability of purified monoclonal or polyclonal mabs, and their ability to tolerate harsh regeneration conditions, make them suitable reagents for non-covalent immobilization. Typically there is a need to immobilize a number of related ligands, with an invariant and a variant portion. If several antibodies against the invariant portion are available, it is likely that at least one of these will be suitable for indirect coupling. A basic approach to developing such a method is described in Protocol

18 Protocol 6. Developing a new antibody-mediated indirect coupling method 1. Obtain a panel of antibodies that can bind a suitable ligand and are available in pure form. 2. Analyse the binding properties and select a high affinity antibody with a slow k off. 3. Covalently couple the antibody by amine coupling (Protocol 4) and check for activity. If inactive, try a different antibody. 4. Establish regeneration conditions (Protocol 5). 5. If suitable regeneration conditions cannot be found, try a different antibody Activity of immobilised ligand It is important to evaluate the functional integrity of the immobilised protein. This is best achieved by using a protein which binds to correctly-folded ligand. An ability to bind its natural ligand is reassuring evidence that an immobilised ligand is functionally intact. Since monoclonal antibodies (mabs) usually bind to 'discontinuous epitopes' on a protein, they are excellent probes of protein structure. It is preferable to check the binding of several mabs, including ones that bind to the same binding sites or, failing this, the same domain of the natural ligand. MAbs are particularly useful if the natural binding partner has not been identified and/or a candidate binding partner being assessed for an interaction with the ligand. In is important to know not just whether immobilised ligand is 'active' but also what proportion is active (Protocol 7). Protocol 7. Quantitating binding levels The ligand activity (ActL), stoichiometry (S), molecular mass of ligand (M L ) and analyte (M A ), and the analyte binding level at saturation (A) are related as follows: ActL % S + M, ) * M L % A ( & ' A L

19 Where L is the level of ligand immobilised S is the molar ratio of analyte to ligand in analyte/ligand complex The product ActL*S is readily calculated once A and L have been measured. Either ActL or S need to be independently determined in order to calculate the other. A convenient way to measure ActL by SPR is to use Fab fragments of mabs specific for the ligand. Intact mabs are less useful because of the uncertaintly as to their binding stoichiometry Control surfaces A control surface should be generated which is as similar as possible to the ligand surface, including similar levels of immobilisation. This is to measure non-specific binding and to record the background response. The immobilisation levels need to be similar because this affects the background response measured with analytes that have a high refractive index (Section 5.3) Re-using sensor chips Because each sensor chip has several flow cells, it is common to have unused flow cells at the end of an experiment. In addition many covalently coupled ligands are very stable, enabling surfaces to be re-used over several days. It is therefore convenient to be able to remove and reinsert sensor chips in the BIAcore (Protocol 8). Protocol 8. Reusing sensor chips "# Undock the sensor chip, choosing the empty flow cell option in the dialogue box. "# Store the sensor chip in its cassette at 4ºC. "# When it is to be re-used, re-insert and dock the sensor chip. "# After priming the system check the 'dips' (Section 3.3)

20 5. Analyte The extent to which the analyte needs to be characterised depends on the nature of the experiment. Quantitative measurements (Section 7) require that the analyte is very well characterised and of the highest quality. In contrast, this is less important for qualitative measurements (Section 6) Purity, activity and concentration In order to determine affinity and association rate constants it is essential that the concentration of the injected material is known with great precision. The only sufficiently accurate means of measuring concentration is to use a spectrophometer to measure the optical density of a solution of pure protein, usually at 280 nm (OD 280 ). In order to calculate the concentration from the absorption at OD 280, two additional measurements are required. Firstly, it is necessary to determine the extinction coefficient. Although this can be calculated from the primary sequence it is best determined directly by amino acid analysis of a sample of the protein with a known OD 280. Secondly, it is important to assess what proportion of the purified protein is 'active', i.e. able to bind to the ligand. This can be done by depletion experiments in which the ligand-coated sepharose beads are used to deplete the analyte from solution (16). mab-coated sepharose beads can be used instead if ligand-coated beads are impractical. If all the analyte can be depleted in this way it is 100% active Valency In general, affinity and kinetic measurement require that each analyte molecule has a single binding site, i.e. is monovalent. If the protein has a single binding site it is only necessary to show that it exists as a monomer in solution. This is most readily achieved by size-exclusion chromatography (Chapter 3) or analytical ultracentrifugation (Chapters 4 and 5). It is important to emphasise that

21 these analytical techniques will not detect the presence of very low concentrations of multivalent aggregated material (17,18). In order to ensure that this material does not contribute to the binding it is ESSENTIAL to purify the monomeric peak by size exclusion chromatography immediately before BIAcore experiments and to analyse it before concentrating or freezing it. Only when it has been shown that concentration, storage or freezing do not affect the measured affinity and kinetic constants is it wise to deviate from this strict principle. Fortunately the presence of multivalent aggregates is readily excluded by analysis of the binding kinetics (Section 7.2). If the dissociation of bound analyte is monophasic (mono-exponential) multivalent binding can be ruled out. If dissociation is bi-exponential with >10 fold difference in the two k off values, multivalent binding is likely. Bi-exponential dissociation with smaller differences in the two k off values could have several explanations (see Section 7.2) Refractive index effect and control analytes When an analyte is injected over a surface it is important, for two reasons, to perform a second injection with a control analyte. Firstly, this helps rule out non-specific binding. Secondly, it controls for any refractive index artefacts. These occur when the background signal measured during the injection of an analyte sample differs between flow cells. Clearly such an artefact will create problems for affinity measurements. It can occur whenever there is a substantial difference between surfaces. For example, if very different levels of material are immobilised on each surface. In this case, because the immobilised material displaces volume, the volume accessible to the injected analyte sample will differ between flow-cells. If the analyte sample has a higher refractive index than the running buffer, a larger background signal will be seen from the surface with less immobilised material. This artefact is greater when (i) there are big differences in the levels of immobilised material (e.g. >2000 RUs), (ii) the background signal is very large (e.g. >1000 RUs), and (iii) the

22 binding response is much smaller (<10%) than the background response. These conditions are common when measuring very weak interactions, because high concentrations of analyte are injected, and with low molecular weight analytes, which give a very small response. A refractive index artefact can be detected by injecting a control solution with a similar refractive index to the analyte sample. If the control solution gives the same response in both flow-cells a refractive index artefact can be excluded. Refractive index artefacts are most easily avoided by taking care to immobilise the same amount of total material on both the control and ligand sensor surfaces Low molecular weight analytes Recent improvements in the signal to noise ratios of BIAcore instruments have enabled binding to be detected of analytes with M r as low as 180 (19). There are two major problems associated with such studies. Firstly, very high densities of ligand must be immobilised in order to detect binding. The levels can be calculated using equation in Protocol 7. For an analyte of M r ~200 that binds a ligand with an M r of ~40000, approximately RU of active ligand needs to be immobilised to see 50 RU of analyte binding. Achieving this level of immobilization is very difficult. A second problem is that with such small binding responses refractive index effects become significant. The latter can be avoided by dissolving and/or diluting the analyte in the running buffer and using a control flow cell with very similar levels of immobilization. 6. Qualititative analysis; do they interact? 6.1. Positive and negative controls The main purpose of a qualitative analysis is to establish whether or not there is an interaction between a given analyte and ligand. If binding is detected it is necessary to test negative controls to exclude a false positive. These include negative ligand controls and negative analyte controls

23 Blocking experiments, using molecules known to block the interaction, are also useful negative controls. In contrast, if no binding is detected it becomes necessary to run positive controls to establish whether this reflects the absence of an interaction, a very low affinity, or an artefact resulting from defective ligand or analyte. The ligand is readily assessed by showing that it can bind to one or more mabs or additional analytes should they exist. The analyte can be tested by injecting it over a surface with either a known interacting ligand or or mab immobilised to the surface Qualitative comparisons using a multivalent analyte The binding of a multivalent analyte can be heavily influenced by the level of immobilised ligand, since the latter will influence the valency of binding. Thus when comparing the binding of a multivalent analyte to different ligands it is important that these ligands are immobilised at comparable surface densities. 7. Quantitative measurements Quantitative measurements are far more demanding than qualitative measurements because of the quality and amount of materials required and the difficulties associated with designing the experiments and analysing the data. When undertaking these measurements it is particularly important to understand the various pitfalls and how these can be avoided (4,5,8). Any quantitative analysis on the BIAcore requires that the analyte binds in a monovalent manner. Since many binding parameters are temperature dependent it is important to perform key measurements at physiological temperatures (i.e. 37-C in mammals). Considering how easy it is to regulate the temperature on the BIAcore it is surprising how seldom this is done

24 7.1. Affinity Concepts As has been described earlier in this volume (see, e.g. Chapters 1, 3-5), there are a number of ways to represent the affinity of an interaction. "# The association constant (K A ) or affinity constant is simply the ratio at equilibrium of the product and reactant concentrations. Thus for the interaction A + B. AB K A CAB, C % C A B Note that K A has units M -1 (i.e. L.mol -1 ) "# Many prefer to express affinity as the dissociation constant or K D, which is simply the inverse of the K A, and therefore has the units M. "# Affinity can also be expressed as the binding energy or, more correctly, the standard state molar free energy (/G-). This can be calculated from the dissociation constant as follows. where K D / G-, RT ln o C T is the absolute temperature in Kelvin ( K = 25 -C) R is the Universal Gas Constant (1.987 cal.k -1.mol -1 ) C o is the standard state concentration (i.e. 1 M) Experimental design In principle the affinity constant can be measured directly by equilibrium binding analysis, or calculated from the k on and k off. However, because of the difficulties associated with obtaining definitive kinetic data on the BIAcore, equilibrium binding analysis is more reliable. It involves injecting a series of analyte concentrations and measuring the level of binding at equilibrium. The

25 relationship between the binding level and analyte concentration enables the affinity constant to be calculated (7). A basic approach to such measurements is outlined on Protocol Preliminary steps Because equilibrium measurements are unaffected by mass-transport or re-binding artefacts, high levels of immobilisation can be used to increase the binding response. This is particularly useful when the background signal is high or the analyte very small. Ideally, equilibrium affinity measurements require that the level of active ligand on the surface is the same for each concentration of analyte injected. This is usually straightforward when the ligand is covalently coupled (and so does not dissociate) and the analyte dissociates spontaneously within a few minutes (so that regeneration is not required). In cases where regeneration is required it must be shown that ligand activity is unaffected by repeated regeneration. Where captured ligands dissociate spontaneously or require regeneration, it may be difficult to maintain the level of the ligand constant. It is possible to correct for this if the level of active ligand can be accurately monitored (20). It is important to ensure that the analyte injections reach equilibrium. While the approximate time it takes to reach equilibrium can be calculated (see footnote to Protocol 9), it is advisable to measure this directly in preliminary experiments under the same conditions (flow rate, analyte concentration, ligand density) as those to be used for the affinity measurements. Enough time must be allowed following the injection for the bound analyte to dissociate completely from the sensor surface (see footnot to Protocol 9). If dissociation is incomplete, or takes too long, it may be necessary to enhance dissociation by injecting regeneration solution The experiment Ideally the analyte concentration should be varied over four orders of magnitude, from 0.01* K D to 100 * K D. However it is often only practical to vary the concentration over 2-3 orders of magnitude

26 This can be achieved with 10 two-fold dilutions starting at between 10 * K D and 100 * K D. An assumption in these affinity measurements is that the level of active immobilised ligand remains constant. This should be checked by showing that a reference analyte binds to the same level at the beginning and end of the experiment. A second internal control is to reverse the order of injections. An efficient way of doing this is to work up from the lowest concentration of analyte, give one injection at the highest concentration, and then work back down to the lowest concentration. The same affinity should be obtained irrespective of the order of injections Data analysis In order to derive an affinity constant from the data a particular binding model must be used. The simplest (Langmuir) model (A+L.AL) is applicable in the vast majority of cases. It assumes that the analyte (A) is both monovalent and homogenous, that the ligand (L) is homogeneous, and that all binding events are independent. Under these conditions data should conform to the Langmuir binding isotherm, Bound C, C A A * Max 0 K D where "Bound" is measured in RUs and "Max" is the maximum response (RUs). C A is the concentration of injected analyte and K D is in the same units as C A (normally M) The K D and Max values are best obtained by non-linear curve fitting of the equation to the data using a suitable computer software such as Origin (MicroCal) or Sigmaplot. A Scatchard plot of the same data (see Chapter 3), obtained by plotting Bound/C A against Bound, is useful for visualising the extent to which the data conform to the Langmuir model. A linear

27 Scatchard plot is consistent with the model. Scatchard plots alone should not be used to estimate affinity constants since they place inappropriate weighting on the data obtained with the lowest concentrations of analyte, which are generally the least reliable. Non-linear Scatchard plots indicate that the data do not fit the Langmuir model. Before considering models that are more complex it is important to exclude trivial explanations (Section 7.1.5) Controls Several artefacts can result in erroneous affinity constants. These include an effect of ligand immobilisation on the binding, an error estimating the active concentration of analyte, and an incorrect assumption that the analyte is monovalent. The most rigorous control is to confirm the affinity constant in the reverse or 'upside down' orientation since this excludes all three artefacts. If this is not possible the experiment should be repeated with the ligand immobilised in a different way, which addresses the possible effects of immobilization on binding. The affinity should be also be confirmed with two independently-produced batches of protein and with different recombinant forms of the same proteins Non-linear Scatchard Plots A non-linear Scatchard plot indicates that binding does not conform to the Langmuir model. Many binding models can be invoked to explain non-linear Scatchard plots. Distinguishing between these models can be very difficult and is beyond the scope of this Chapter. The priority should be to exclude trivial explanations for non-linear Scatchard plots. A 'concave up' Scatchard plot is the most common deviation from linearity. It may be a consequence of heterogeneous ligand, multivalent analyte, or (rarely) negative cooperativity between binding sites. A trivial cause of analyte heterogeneity is the presence of multivalent analyte. Ligand heterogeneity may be a consequence of

28 immobilisation. A 'concave down' Scatchard plot is unusual and indicates either positive cooperativity between binding sites or self-association of the analyte, either in solution or on the sensor surface. The control experiments outlined in will help to eliminate some trivial explanations. For example, if the analyte has a multivalent component the non-linear Scatchard plot will not be evident in the reverse orientation. The shape of the Scatchard plot will also depend on the surface density of ligand. If the ligand immobilization is responsible for heterogeneity, this should be eliminated in the reverse orientation and if the ligand is immobilised indirectly. Protocol 9. Affinity measurements Preliminary steps 1. Immobilise the ligand and a control. High levels of immobilisation are acceptable. 2. Ensure that the analyte is monomeric and binds monovalently. Determine accurately the concentration of the analyte and the proportion that is active. 3. Determine the time it takes to reach equilibrium and the time it takes for the bound analyte to dissociate completely from the sensor surface. While this should be done empirically, under the conditions to be used for the equilibrium measurements, approximate times can be calculated from the k a off. 4. If necessary, establish regeneration conditions (Protocol 5). 5. Obtain a rough estimate of the K D by injecting a series of five-fold dilutions. Measurements 1. Prepare a dilution series of analyte starting at times the K D with at least 9 two-fold dilutions thereof. There should be enough for two injections at each concentration except the highest concentration, where only enough for one injection is required. A minimum of 17!L of sample is required per injection b. 2. Make up separate control sample of analyte (at concentration ~ K D ) with enough for 2 injections. 3. Set the flow rate. To conserve sample this can be as slow as 1!L.min

29 4. Inject the control sample. 5. Inject the dilution series starting from low and moving up to the highest concentration (low-tohigh) and then moving back down to the low concentrations (high-to-low). Inject the highest concentration only once. It is important to inject for a period sufficiently long to reach equilibrium. Either enough time must be allowed for spontaneous dissociation or the analyte must be eluted with regeneration buffer. 6. Repeat the injection of the control sample. 7. For all injections measure the equilibrium response levels in the ligand and control flow cells. The difference between these two is the amount of binding at each concentration. Data analysis 1. Plot the binding versus the concentration for both the low-to-high and the high-to-low series (20). 2. Fit the Langmuir (1:1) binding isotherm to the data by non-linear curve fitting. Use this to determine the K D and maximal level of binding. 3. Do a Scatchard plot and check if it is linear. The points on the plot where binding is less than 5% of maximum are highly inaccurate and should be ignored. 4. If the Scatchard plot is not linear do further experiments to establish cause (Section 7.1.5). Controls 1. The affinity constant should be confirmed in the reverse orientation. 2. Should this be impossible (e.g. because the ligand is multivalent) the affinity constant should be confirmed with the ligand immobilised by a different mechanism, preferably by ligand capture. 3. Use at least two independently produced batches of protein. 4. Use different recombinant forms of the same proteins. a Both the time taken to reach equilibrium and the time it takes for the bound analyte to dissociate are governed primarily by the k off. For the simple 1:1 model binding will reach 99% of the equilibrium level within 4.6/k off seconds. Similarly, it will take 4.6/k off seconds for 99% of the analyte to dissociate. Thus for k off ~0.02 s -1, equilibrium will be reached within ~ 230 s and the bound analyte will take ~230 s to dissociate. b If the Quickinject command is used as little as 15!L of analyte sample is used